Skip to content

PyTorch implementation of BCQ for "Off-Policy Deep Reinforcement Learning without Exploration"

Notifications You must be signed in to change notification settings

agarwl/off_policy_mujoco

 
 

Repository files navigation

REM + TD3 (Doesn't work)

- Note this is a preliminary repository and *no effort* was spent getting REM to work well with TD3 (as it doesn't work).
- This repository should should not be used for benchmarking and only be used a starting point for MuJoCo experiments.

If you use this code, please cite the the paper. To launch batch experiments with RSEM, use the file run_main.sh. To generate data, use run_expert.sh. RSEM works somewhat but not well though.

Method is tested on MuJoCo continuous control tasks in OpenAI gym. Networks are trained using PyTorch 0.4 and Python 2.7.

About

PyTorch implementation of BCQ for "Off-Policy Deep Reinforcement Learning without Exploration"

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 97.0%
  • Shell 3.0%